Cortical Maps of Resistive Anisotropic Networks


Work Area: Basic Aspects of Multiple Computing Agents

Keywords artificial vision systems, neural networks, architectural specifications

Start Date: to be announced / Status: starting

[ participants / contact ]

Abstract The project will investigate how the information processing capabilities of neural networks can be exploited in the direct (VLSI) implementation of complex algorithmic tasks such as those occurring in natural and artificial visual systems. Research will address: (i) the computational modelling of natural systems as formal neural networks; (ii) the formalisation of neural computation through an abstract hierarchy of operators both for low- and high-level capabilities; (iii) the architectural specification of networks of these operators, specified as multilayer meshes of simple processing elements; (iv) the methodological issues related to the programmability/design of these meshes specified at the circuit level.


The project aims to investigate the information processing capabilities of hierarchically structured networks of multiple and cooperating simple agents. The relationships amongst natural, formal, and artificial neural networks will be considered. Specifically, the project will address the types of information processes characterising low- and intermediate- level tasks of visual perception. The computational principles of these networks are also applicable to other complex information processes encountered in perceptual tasks.

Approach and Methods

The following assumptions will guide our investigation: (i) visual information processing can be expressed through a set of processing operators acting on the image; (ii) it is possible to interpret the processing operators in terms of simple computing agents of a network (such as a mesh of (active) resistors or a neural biological network); (iii) the processing operators are organised in complex, hierarchically structured layers. These points shall be addressed in strong cooperation with computational neuroscience and physiology. This will be done by considering the links between single artificial operators and physiologically observed receptive fields and how network architectures correspond to realistical topographical brain maps. The project gathers expertise and knowledge from different disciplines: Applied Mathematics, Electrical Engineering, Information Science, Neuroscience.


We anticipate this effort to bring convincing results to the industry and to raise the interest for new types of computing architectures that could lead to innovative products. Various applications will benefit from this research: image processing, optical character recognition, speech recognition, handwriting recognition. We want to open new perspectives on the conception, design and possible implementation of this type of high performance computing system.


Universita di Genova - I
DIBE - Department of Biophysical and Electronic Engineering
Via all'Opera Pia 11a
I- 16145 GENOVA


Ruhr Universität Bochum - D
University of Bonn - D


Prof. G.M. Bisio
tel +39-10 3532756
fax +39-10 3532795

LTR synopses home page LTR work area index LTR acronym index LTR number index LTR Projects index
All synopses home page all acronyms index all numbers index

CORMORANT - 8503, August 1994

please address enquiries to the ESPRIT Information Desk

html version of synopsis by Nick Cook